It' difficult to help without any hint, but probably you'll have to convert the image data to your training data format. Something like:
vector<vector<uchar>> training_data;
vector<int> training_labels;
vector<Mat> images;
for(int i=0;i<images.size();i++){
vector<uchar> data;
for(int y=0;y<images[i].rows;y++){
uchar *p=images[i].ptr(y);
for(int x=0;x<images[i].cols;x++){
data.push_back(p[x]);
}
}
training_data.push_back(data);
}
mynetwork.train(train_data,train_labels);
The code is for you to get an idea how to transform your images into a training set; you'll have to adapt it to the data format needed by your network.
we cannot help you, unless you get more specific. show us your code, tell us, where the exact problem is, please.